Top
Best
New

Posted by coloneltcb 1 day ago

HN Slop: AI startup ideas generated from Hacker News(www.josh.ing)
227 points | 95 comments
wanderingstan 1 day ago|
Feature request: option to select a year so you can get retro startup ideas from e.g. web 2 or crypto-mania eras!
Bjartr 23 hours ago||
> DocuQuest: A platform that leverages LLMs to transform and simplify complex technical documentation into interactive, user-friendly learning experiences tailored for developers and engineers.

So "A Young Lady's Illustrated Primer" from Diamond Age, but for devs. A neat idea!

In general docs ecosystems tend to be heavy on only one of reference / explanation / tutorial. Would be cool to have a way to write one and get the others.

Jarwain 3 hours ago||
Reminds me of this project

A game of learning your homelab into a cyberpunk mystery adventure | Hacker News https://share.google/WedMuRgx5WgreNqSN

its_down_again 18 hours ago|||
Actually this sounds great. I got way more out of codecademy’s in-browser, interactive challenges than I did in my middle & high-school classes programming classes. The "learn by doing" process really built my confidence. If you could "demo" your docs directly in the browser it's much easier to learn by doing. I think that'd drive up adoption and you might even crowdsource bug discovery.
mathiaspoint 20 hours ago||
Publishing docs and example project and letting them get scraped into LLMs already accomplishes this pretty effectively.
BlindEyeHalo 1 day ago||
Claude API error: {"type":"error","error":{"type":"invalid_request_error","message":"Your credit balance is too low to access the Anthropic API. Please go to Plans & Billing to upgrade or purchase credits."}}
throwanem 23 hours ago||
SlopCharger: Broker free-tier accounts with fly-by-night AI providers cross-matched with the latest in prompt exploits to enable new billing efficiencies by running your HN AI shitpost on someone else's one-fiftieth of a dime! :sparkles:
kevindamm 20 hours ago||
EdgeSlop: now you can bring your own API key for a wider range of choices -- BYOAI!
jshchnz 21 hours ago|||
Sorry about that, was sleeping and this got on the front page again, just upped the credit balance to fix!
jerf 22 hours ago|||
Really should have been some caching involved. We didn't all need our own personal slop.
jshchnz 18 hours ago||
Jerf, everyone deserves their own personal slop
ActionHank 22 hours ago||
This is poetic, because AI will become inaccessible pricing-wise and all these AI slop shops are going to have a hard time keeping customers.
vntok 19 hours ago||
How do you see AI becoming inaccessible pricing-wise? If anything it seems models are measurably more performant and measurably more efficient every month? People regularly run huge models locally on consumer hardware, which used to be unfathomable months ago..
amelius 1 day ago||
This reminds me of https://www.halfbakery.com/ which was much more fun, probably because the ideas came from actual humans.
throwanem 23 hours ago|
Technically, so do these.
amelius 21 hours ago||
Ok, "came directly from humans" then.
jmknoll 22 hours ago||
Alright, but DocuQuest is actually a really good idea.

"DocuQuest: A platform that leverages LLMs to transform and simplify complex technical documentation into interactive, user-friendly learning experiences tailored for developers and engineers."

WA 22 hours ago|
I recently found a vibe-coded app that generates courses on the fly from YouTube, including automatically generated quizzes. I forgot the URL, but the result was beyond awful. Questions were something like: "what was the title of the youtube video?" and other utterly stupid things.

Not saying that it isn't possible, but stuff like this does need the human touch.

trevor-e 1 day ago||
>SciDigest: An AI-powered platform that transforms complex scientific research into engaging, digestible content for the general public, making scientific knowledge accessible and actionable.

This would actually be great. So many researchers have a marketing problem with explaining and getting people excited for their work.

thicknavyrain 1 day ago||
I'm a popular science writer with eight year's experience doing exactly this (SciShow, Crash Course, Veritasium and recent winner of the Wellcome Collection Non Fiction Awards) without AI. Done right, the right coverage of even a pre-print reached hundreds of thousands/millions of people. But I've experimented with every SOTA model since 2022 with the most detailed and specific prompting I can think of (including multiple examples of transcripts of work already in the public domain) to see if it can replicate good quality science communication.

The content is usually reasonably strong but the tone is always off and it never quite understands what it is a reader/viewer needs to really get to grips with the topic if they don't already have a prior foundational understanding (though I notice this about a lot of other media outlets with professional science communicators too). It also has poor editorial thinking around what bits are most likely to be interesting and cohesive when considered as part of the whole piece.

But I'm still reasonably convinced as AI improves it ought to be able to replace me with the right workflow/context/prompting. I think there will always be a demand for my (and many other writers') talents as they are so it doesn't really bother me, but it'd be great to extend the work to all the many scientific discoveries that don't get the same attention. If anyone is serious about developing something like this, I'd be interested in partnering with them as someone with domain expertise on science communication and familiar with prompt engineering (email in bio).

trevor-e 1 day ago|||
That's super cool, I love the SciShow videos.

I think you're right about the editorial thinking + what do people find interesting parts. But that doesn't have to be solved by directly by AI, it's easy enough to sidestep the problem and provide a nice interface for the human-in-the-loop part. I'd imagine that would save you a ton of time by having a nice starting point depending on how much you have to rewrite for tone.

thicknavyrain 1 day ago||
That's true, it could just turn the writer's role into more of an editorial role. The main time-saving I have so far is being able to upload papers and get it to fact check for me. The editorial guidelines at SciShow are stricter than any academic journal I've published in: any non-trivial statement has to be supported by a direct, findable quote in (most-of-the-time) peer-reviewed scientific literature. I once had to find a citation for the idea that heat + fuel + oxygen generates a fire! (for this video: https://www.youtube.com/watch?v=BEcaE0e0CZg)

LLMs make that much easier. As I collect primary sources during my drafting/writing phrase, I can type up any non-trivial claims I'm making in my script in a separate document, share that with the LLM and say "Quoting directly from the set of attached PDFs, identifying which document, and on which page the quote comes from, find content which directly supports each of these assertions" and it generally goes a great job. At any rate, I have to check each of those quotes for accuracy but the help in _finding_ those quotes in order to pass a stringent fact checking procedure is a huge help if I didn't scribble down the supporting quotes during my research phase. This is also, by the way, stricter than the fact checking process for most non-fiction publishing.

thekracken 1 day ago||
>SciShow are stricter than any academic journal I've published in

Now there's a testimonial. I look forward to browsing the source links with each video!

notahacker 1 day ago||||
Feels like there might be an accuracy issue as well. Although that might make it perfectly suited to replacing whoever writes university press releases...
kristianp 9 hours ago||||
> as AI improves it ought to be able to replace me with the right workflow/context/prompting

The bitter lesson may very well come for us all...

thekracken 1 day ago||||
>I've experimented with every SOTA model since 2022

>The content is usually reasonably strong but the tone is always off and it never quite understands what it is a reader/viewer needs

A SOTA model fine-tuned with your choice of transcripts could probably get you most of the way there. There might be a customized, open-weight model already on Huggingface that meets your needs.

antonvs 1 day ago|||
One question is whether the audience is discerning enough to care about the issues you're raising. This seems like it could be a variation of why the umpteenth Marvel movie beats out indie masterpieces at the box office. The audience for high quality becomes increasingly niche as the market's relatively low bar for quality is satisfied.
janpmz 1 day ago|||
I've attempted this with pdftomp3.com where you can listen to PDFs. It has an "AI Explanation" mode where the content of the PDF gets explained. Its like a NotbookLM podcast, but I was earlier :)

Currently I'm working on an app for that, because thats where I listen to the MP3s anyways.

falcor84 1 day ago|||
I've been using the NotebookLM "podcasts" for this. I upload an arxiv pdf and use the Interactive Mode to have them talk about it, while I can pause them to ask clarifying questions. It's been surprisingly efficient for me to get an initial grasp on things I'm not familiar with, at times prompting me to then go on exploring interesting rabbit holes.
bookofjoe 23 hours ago|||
https://www.comunicarestiintifica.ro/en/category/scidigest/
dax_ 1 day ago|||
I'm 99% sure I already saw a product launch on HN for precisely this idea.
IsHeInFarside 1 day ago||
I am trying to do this, trying to generalize a workflow that at least has helped me grok a few papers.
davidjhall 23 hours ago||
Very cute but wish there was a filter for the flavor of the week ... I feel 60% of them will have LLM because of this bias. Last month would've been "powered by rust."
mittermayr 1 day ago||
This should be connected to https://comsensei.com to get an available domain for each one of the ideas. It's all coming together :)
nusl 1 day ago||
You're promoting your own website based on AI-generated things on a Hacker News post aimed at generating AI slop.
mittermayr 1 day ago|||
I've received nothing but nice messages about the site after posting it and people seem to be getting a kick out of it. I feel like this (if barely so) meets the bar to give it a pass perhaps, especially considering that my voice is worth as much as yours and the next person, so without upvotes, it will disappear, just how it's meant to be. Hope that helps a bit.
mouse_ 1 day ago|||
good for him i hope he makes it
fennecbutt 1 day ago||
Where do we draw the line at self promotion (advertising) then?
nine_k 21 hours ago|||
If anything, this was a very short and quite contextual piece of text-only advertising. If all online advertising were like that, I won't have to run an ad blocker.
collingreen 20 hours ago||||
For me, undisclosed self promotion puts a very bad taste in my mouth, to the point of coming off as bad faith. I dislike pretending you're just mentioning a cool service since you're effectively trying to trick me with false social proof.

If the post had said "I made xyz which could auto generate domains for these ideas using ai to fully close the slop gap" I wouldn't mind and might even appreciate the additional fun.

mouse_ 1 day ago||||
As long as it's not apparently being done by a bot, who cares?

This is effectively an advertising website.

haswell 1 day ago|||
This is not primarily an advertising site. See guidelines:

> Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.

nmilo 21 hours ago||
> It's ok to post your own stuff part of the time
haswell 18 hours ago||
Yes. I was responding specifically to this claim:

> This is effectively an advertising website.

Which remains untrue even if you're allowed to post your own stuff from time to time.

mrd3v0 1 day ago||||
> not apparently being done by a bot

LLMs will make this very hard to detect, very soon if not already.

iJohnDoe 20 hours ago|||
Most of the time, the top post of a Show HN is someone else talking about their own project or something off-topic from the Show HN.

Self-promotion on HN is constant.

vntok 19 hours ago|||
No need to do that. The provenance does not matter as long as the content is interesting, which is the case here.

This is the original hacker's ethos: put something out there for others to use, and let the people make their own opinion about the thing itself, not the authoritativeness of its source.

oceanhaiyang 23 hours ago||
I like how I submit a request and it never loads.
unleaded 20 hours ago||
I miss that small period of AI where it excelled at making random shit in this vein( see https://x.com/dril_gpt2 and https://www.reddit.com/r/SubSimulatorGPT2/top/?t=all [somewhat nsfw text]).

I'm not really sure why modern AI can't really do stuff like that anymore—I would guess a combination of being whacked with a crowbar to submit to humans (RLHF, but I'm not sure if it affects base models?), alignment stuff, and just being too smart.

Would love to see something like this with GPT-2 and how it compares

anitil 8 hours ago|
If you are looking for an offline solution, one I came across recently as a python one-liner (requires faker)

  python3 -c 'print(__import__("faker").Faker().bs())'
More comments...